Skridlov
Posts: 116
Location: United Kingdom
Registered: 12 Nov 2007
|
"Nodal" point precision.
Posted: 18 Feb 2008 at 12:35 GMT
|
|
I have been taking a look at the settings I have been using. The configuration is D200; 10.5 Nikkor; NN3 (on a Manfrotto 338 levelling plate). The NN site has values for this lens/camera combo of 53mm lower, and 80mm upper arms. I checked this using Rosario's "oven grill" method and discovered a noticeable degree of parallax error. Adjusting the upper value to 79mm and then 81mm both worsened the problem slightly.
Shooting perpendicularly down onto the panhead levelling plate with two shots at 180 deg separation and then overlaying the two shots in CS2 reveals substantial shift. When looking through the viewfinder in both positions the centre of the shot looks substantially identical (although I relise that the combination of the focal length and eye/viewfinder parallax probably makes this a meaningless statement).
I'd be interested to hear what anyone else has noticed with this common equipment configuration. Assuming that it's impossible to get it 100% perfect, what is acceptable?
|
|
alert moderator
|
|
Matt Rogers
[360 Precision]
Posts: 198
Location: Oxford UK, United Kingdom
Registered: 16 Jun 2005
|
Re: "Nodal" point precision.
Posted: 18 Feb 2008 at 12:51 GMT
|
|
Skridlov said: Shooting perpendicularly down onto the panhead levelling plate with two shots at 180 deg separation and then overlaying the two shots in CS2 reveals substantial shift. When looking through the viewfinder in both positions the centre of the shot looks substantially identical (although I relise that the combination of the focal length and eye/viewfinder parallax probably makes this a meaningless statement).
You will never get two opposing images shot at 180º apart to align in Photoshop due to the camera's sensor offset. Using the view finder to align the central axis of the camera is a mistake. What you need to do is measure the distance from the base of the camera to the centre of the lens mount. You then need to work-out the relationship between the setting on your panohead and this measurement to be able to accurately set the camera position.
On the 360Precision Adjuste for example the setting on the base slider corresponds directly to the base to lens mount measurement. This allows you to quickly and accurately set your cameras central alignment point.
You should be able to set the no parallax point to within 1/4 pixel (12mp image) or around 0.25mm on your panohead. If set correctly you should be able to consistently achieve average control point distances under 1 pixel over a 10,000 x 5,000 pixel panorama.
Matt
|
|
alert moderator
|
|
Skridlov
Posts: 116
Location: United Kingdom
Registered: 12 Nov 2007
|
Re:
Posted: 18 Feb 2008 at 13:05 GMT
updated: 18 Feb 2008 at 13:15 GMT
|
|
Hi Matt. Thank you for this information. All I have to do now is digest it!
You say:
"What you need to do is measure the distance from the base of the camera to the centre of the lens mount. You then need to work-out the relationship between the setting on your panohead and this measurement to be able to accurately set the camera position."
I'll have to think about this and get my hands on the camera to understand what the implications are. meanwhile I'm quite surprised that I haven't stumbled across this before. It's a very common hardware configuration so I'm surprised I haven't come across the applicable values somewhere!
Afterthought. Does this mean that the position of the camera has to be adjusted betwen the two shots or is there a mean position that elliminates it?
|
|
alert moderator
|
|
Hans Nyberg
Posts: 765
Location: Denmark
Registered: 28 Aug 2005
|
Re: Re:
Posted: 18 Feb 2008 at 14:57 GMT
|
|
Sensorshift is a very common thing and it differs from camera to camera.
My old D60 and 10D had a lot. My 20D is perfect and my 5D is almost perfect.
Here is an illustration from the 10D.
The Sigma image circle has the centre at the red point which has both an horisontal and a vertical shift from the sensor centre.
When you turn the camera 180 degrees this will of course shift the center to the other side.
Seeing this with a fullframe image, fisheye or rectilinear is impossible.
You can see it on a circular fisheye.
Hans
|
|
alert moderator
|
|
Skridlov
Posts: 116
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: Re:
Posted: 18 Feb 2008 at 15:30 GMT
|
|
|
Hans - my question, unsurprisingly, is how did you accommodate the error with your previous cameras?
|
|
alert moderator
|
|
gus
Posts: 378
Location: United Kingdom
Registered: 19 Jun 2007
|
Re: Re:
Posted: 18 Feb 2008 at 17:36 GMT
|
|
I might be wrong, but once you've saved a template for your lens/camera combination in Ptgui, surely this will take care of the sensor shift? The sensor shift is constant, and since the template was created using well placed control points, the sensor shift becomes irrelevant in future deployment of the same template.
That's my uneducated guess anyway…...
|
|
alert moderator
|
|
Skridlov
Posts: 116
Location: United Kingdom
Registered: 12 Nov 2007
|
Re: Re:
Posted: 18 Feb 2008 at 18:55 GMT
|
|
The problem is that I had been hoping to use the technique of making 2 nadir shots at 180 deg off the tripod, layering them with one 180 deg reversed in PS & then painting out the panhead arm. This would leave a relatively small hole to be patched by a section from a hand-held shot.
The error I'm seeing is much bigger than I would have expected.
I've have tried adjusting the position (notionally 53mm according to the NN database)by a couple of mm but it only exacerbates the problem. Not quite sure what to do about this as I don't really understand previous posters' observations about the problem.
|
|
alert moderator
|
|
Shantic
Posts: 274
Location: Puerto Vallarta, Jalisco, Mexico, Mexico
Registered: 12 Apr 2005
|
Re: Re:
Posted: 18 Feb 2008 at 19:06 GMT
|
|
|
Skridlov, I usually when nedded don't open the tripod all the way, depending on the terrain you could do it with the tripod just barely open, this leaves a very small tripod foot-print, and then just use a hand held downshot, photoshop it, and you are set, I also tried the method you want but for me, the one I describe is easier and faster.
|
|
alert moderator
|
|
John Houghton
Posts: 2198
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: Re:
Posted: 18 Feb 2008 at 19:27 GMT
|
|
Skridlov said: The problem is that I had been hoping to use the technique of making 2 nadir shots at 180 deg off the tripod, layering them with one 180 deg reversed in PS & then painting out the panhead arm. This would leave a relatively small hole to be patched by a section from a hand-held shot.
What you need to do is to add an alpha channel to the pair of shots and mask out the pano head and tripod. Then stitch both images in with all the other shots. You will be left with a small hole to fill with a handheld shot.
John
|
|
alert moderator
|
|
Hans Nyberg
Posts: 765
Location: Denmark
Registered: 28 Aug 2005
|
Re: Re:
Posted: 18 Feb 2008 at 20:45 GMT
|
|
Skridlov said: The problem is that I had been hoping to use the technique of making 2 nadir shots at 180 deg off the tripod, layering them with one 180 deg reversed in PS & then painting out the panhead arm. This would leave a relatively small hole to be patched by a section from a hand-held shot.
.
Skridlov
You should try shooting the 6 around at -10 degrees instead.
You save a lot of work with it.
I introduced it 3 years ago when I got my 10,5mm and I use it today with my 5D and a 15mm or the Tokina 10-17mm.
Most VR photgraphers who use a fullframe fisheye use it this way.
Here is one shot at -10 with the 15mm before I added the Nadir shot.
www.panoramas.dk/panorama/nadir/15mmfisheye.html
You get almost the same with the D200 and the 10,5mm
Hans
|
|
alert moderator
|
|
Nick Fan
[NodalNinja]
Posts: 252
Location: Hong Kong
Registered: 26 May 2006
|
Re: Re:
Posted: 19 Feb 2008 at 3:07 GMT
updated: 19 Feb 2008 at 3:10 GMT
|
|
Skridlov said: The problem is that I had been hoping to use the technique of making 2 nadir shots at 180 deg off the tripod, layering them with one 180 deg reversed in PS & then painting out the panhead arm. This would leave a relatively small hole to be patched by a section from a hand-held shot.
To get smallest possible nadir, you should rotate NN3 by 90 degrees, not by 180 degrees.
you can align the images in photoshop or stitch them in PtGui as John suggested.
you can also tilt the upper rail down by 15 degree. here is a tutorial.
fromparis.com/html/technical_us_quicktime_vr_smal...
nick
|
|
alert moderator
|
|
Tusk
Posts: 20
Location: Dubai, United Arab Emirates
Registered: 14 Feb 2008
|
Re: Re:
Posted: 19 Feb 2008 at 4:54 GMT
updated: 19 Feb 2008 at 5:02 GMT
|
|
|
Hi Nick, I think he means the two shots are at 180deg intervals, aimed straight down towards the tripod (90deg). It's the same technique I have been using & also have a hard time lining them up in Photoshop to erase the arm. I'l have to try the 10-15 deg tilt method, but sounds like it could do some good. I always thought it was absolutely critical to have the cam 100% level on my NN3, & thought that was half the problem.
|
|
alert moderator
|
|
Nick Fan
[NodalNinja]
Posts: 252
Location: Hong Kong
Registered: 26 May 2006
|
Re: Re:
Posted: 19 Feb 2008 at 5:23 GMT
|
|
Tusk said: Hi Nick, I think he means the two shots are at 180deg intervals, aimed straight down towards the tripod (90deg). It's the same technique I have been using & also have a hard time lining them up in Photoshop to erase the arm. I'l have to try the 10-15 deg tilt method, but sounds like it could do some good. I always thought it was absolutely critical to have the cam 100% level on my NN3, & thought that was half the problem.
hi Tusk ,
I know what he means. yes, you get smaller nadir by rotating 90 degrees instead of 180 degrees. Both sides of the lower rail block more view than the rotator base. when rotated by 180 degrees, you get a rectangular nadir, defined by the shorter side of lower rail. When rotated by 90 degrees, you get a circular nadir, defined by the rotator base. of course, I assume the tripod footprint is smaller than the rotator base of NN3. This is not true when a leveler is used.
nick
|
|
alert moderator
|
|
Tusk
Posts: 20
Location: Dubai, United Arab Emirates
Registered: 14 Feb 2008
|
Re: Re:
Posted: 19 Feb 2008 at 6:31 GMT
|
|
Got it - thanks Nick.. I'm definitely off to do a few experiments. It should save quite a bit of time in Photoshop, especially if you are doing multiples for dynamic range.
Regards
Stef
|
|
alert moderator
|
|
Nikos Giannakopoulos
Posts: 60
Location: Athens, Greece
Registered: 10 Jan 2008
|
Re: "Nodal" point precision.
Posted: 19 Feb 2008 at 9:05 GMT
|
|
Matt Rogers said:
You will never get two opposing images shot at 180º apart to align in Photoshop due to the camera's sensor offset. Using the view finder to align the central axis of the camera is a mistake. What you need to do is measure the distance from the base of the camera to the centre of the lens mount. You then need to work-out the relationship between the setting on your panohead and this measurement to be able to accurately set the camera position.
You should be able to set the no parallax point to within 1/4 pixel (12mp image) or around 0.25mm on your panohead. If set correctly you should be able to consistently achieve average control point distances under 1 pixel over a 10,000 x 5,000 pixel panorama.
Matt
That's very interesting information Matt, thanx for sharing it. But is there any illustrated version of this in order to better understand for e.g. which is the "centre of the lens mount" ect.
Thanx in advance for any input.
Friendly Nikos
|
|
alert moderator
|
|
Prev
